Sorry I'm late with the February papers, but preparations for the NYAS 
Symposium on aneurysm disease (Mar 8-9) left me with no time.  
Incidentally, the scientific meeting was a *splendid* success; and the 
book should be the definitive statement as to the state of knowledge on 
the subject in 1996.  Since I have so many other things to catch up on, I 
am taking a short-cut this month and posting a couple of my own 
manuscripts.  The first was published in Biochem Biophys Research 
Communications on Feb 6, 1996; and the other will be published in the 
Proceedings of the NYAS meeting.




PARTIAL AMINO ACID SEQUENCE OF A NOVEL 40 kDa HUMAN AORTIC
PROTEIN, WITH VITRONECTIN-LIKE, FIBRINOGEN-LIKE, AND CALCIUM
BINDING DOMAINS: AORTIC ANEURYSM-ASSOCIATED PROTEIN-40 (AAAP-40)
[HUMAN MAGP-3, PROPOSED]

Shichao Xia, Kathleen Ozsvath, Hitachi Hirose, and M David Tilson*

Department of Surgery, St. Luke's/Roosevelt Hospital Center,
Columbia University, 1000 Tenth Avenue, New York, NY  10583

   -----------------------------------------------------------
Summary:  A microfibrillar protein (40 kDa) purified from the
adventitia of the human abdominal aorta is immunoreactive with IgG
harvested from the wall of abdominal aortic aneurysms.  We have
partially sequenced this protein and found that it has fibrinogen-
alpha-, beta-, and gamma-like domains, a vitronectin-like domain,
and a possible site for binding calcium.  Because of homologies
with other microfibril-associated glycoproteins and because it is
the third member of the family to be characterized in man, we
suggest the name MAGP-3.
   ----------------------------------------------------------

     Nomenclature of the microfibrillar proteins associated with
the elastin fiber is confusing.  A principal component of the
microfibril is fibrillin (fib-15), discovered by Sakai et al, and
Marfan's syndrome has been traced to mutations in the gene for
fibrillin on chromosome 15.    A bovine microfibrillar protein
(Mr ~31 kDa) was discovered in 1986 by Gibson et al,  who coined
the term "microfibril-associated glycoprotein" (MAGP).   Bashir et
al have also cloned the gene for this protein.  Kobayashi et al
reported a 36 kDa calcium-binding protein, also in cow, with tissue
distribution uniquely limited to the aorta (MAGP-36).  The second
human MAGP (deduced MW 21 kDa) was recently reported to have an
open reading frame of 255 amino acids and to be linked to Smith
Magenis syndrome.   The authors of the paper describing the Smith
Magenis protein prefer the abbreviation "MFAP", to avoid confusion
with abbreviations for microfilamentous proteins.  

     We have recently reported that IgG from the aortic wall of
patients with abdominal aortic aneurysms (AAA) is immunoreactive
with a human aortic protein (MW ~ 80) that has features of the
bovine aortic protein of Kobayshi et al (MAGP-36).  MAGP-36
occurs in nature as a disulfide-bonded dimer, so we undertook
further tissue extractions under reducing conditions as described
by Prosser et al.   This approach has led to the partial
characterization of a protein of ~40 kDa that is immunoreactive
with AAA IgG.  We call this protein  Aortic-Aneurysm-Associated
Protein-40 (AAAP-40).  The present communication is to report its
partial sequence and to suggest that, since it is the third human
microfibrillar protein to be described, it be called MAGP-3.

                             METHODS

     Human aortic tissue was extracted for microfibrillar proteins
according to the method of Prosser.11  In brief, the tissue was
first extracted in a phosphate buffer containing potassium chloride
0.6 M.  The insoluble pellet was treated with bacterial collagenase
in Tris buffer.  The final tissue extraction utilized guanidinium
chloride 6 M in buffer containing dithiothreitol 50 mM and EDTA 2
mM.  Gel slices containing the protein of interest were digested
with trypsin or Lys-C and amino acid sequences were determined in
the Protein Chemistry Core Facility, Howard Huges Medical
Institute, Columbia University (New York, NY).

                             RESULTS

     A 59 residue sequence of AAAP-40 as experimentally determined
is shown in Table I.  Alignment on MFAP-4 and MAGP-36 (bovine) is
shown, along with homologous sequences from the alpha and beta
chains of human fibrinogen.  Five and 11 residue sequences of AAA-
40 are shown in Table II, in alignment with sequences from
vitronectin, MFAP-4, MAGP-36, fibriogen-beta (from its calcium
binding domain), and two other calcium binding proteins (myeloid
calcium binding protein and bovine aggrecan).  

                          DISCUSSION

     Fibrinogen-like domains are well-known in the MAGP's.  The
sequence of AAA-40 shown in Table I has  regions of substantial
homology with sequences in the alpha and beta chains of fibrinogen.
Another sequence that we have determined (data not shown) matches
residues 283-292 in the gamma chain.  Since the three fibrinogen
chains are believed to have a single ancestral gene, it would
appear likely that AAAP-40 is related to the common ancestor since
it has motifs that are used in all three fibrinogen subunits.  

     Kobayashi et al noted that MAGP-36 has the property of
calcium-binding, although a candidate site for the calcium-binding
domain has not been proposed.  Kielty and Shuttleworth have
observed that incubation of intact microfibrils with EDTA rapidly
results in gross disruption of microfibrillar organization, which
can be reversed by replacing calcium.  Since fibrillin has 43-
EGF-like motifs with calcum binding consensus sequences, and
calcium has been proposed to orchestrate the assembly of
tropoelastin to the microfibril and hold it in register for
crosslinking, we hypothesize that the calcium-binding domain of
AAAP-40 may play a role in calcium-dependent microfibril assembly
in the aorta.  When we searched GenBank for homologies of AAAP-40
and MAGP-36, we found sequences in calcium-binding myeloid-related
protein (>pir|A44111:# 144-154), the calcium-binding domain of
human fibrinogen-beta, and bovine aggrecan (>pir|A39808: # 59-66)
that have similarities to MFAP-4, MAGP-36, and AAAP-40.  Bold type
is used in Table II to highlight residues that appear to be
conserved, with possible significance for the calcium-binding
function.
     Another matrix protein detected in human embryonic tissue
(sulfated protein 30 kDa=SP-30) has been reported to be
immunoreactive with monoclonal antibodies against human
vitronectin, and it co-distributes in tissue with the protein
that is immunoreactive with antibody against MAGP-31.14  A sequence
of AAAP-40 that matches residues # 230-240 in human vitronectin is
also shown in Table II.  Tomasini-Johansson et al proposed that
SP30 is the human homolog of MAGP-31, but since MAGP-31 does not
have a vitronectin -like domain, we believe that SP-30 is more
likely to be closely related to AAAP-40.  

     Finally, a brief comment on the nomenclature problem.  Perhaps
the simplest approach for the present would be to assign the human
MAGP's a number in the order of their discovery.  Thus, the first
would be the principal component of the microfibril, fibrillin. 
The second would be the protein of Smith Magenis syndrome, and the
third would be the protein described in the present communication. 
MAGP-3 is probably the human homolog of the bovine protein of
Kobayashi (MAGP-36), and the human homolog of Gibson's MAGP-31 has
yet to be identified.  Notwithstanding the proposal of Zhao et al
to call this family of glycoproteins "MFAP",9 we favor retention of
the abbreviation MAGP, since the first bovine member of this family
was so-named by Gibson et al ten years ago and the term has been
widely used ever since.                               TABLE I

     This Table shows sequence of AAAP-40, as determined
experimentally.  The sequence of AAAP-40 is aligned along a
continuous sequence of MFAP-36, beginning at residue 140. 
Homologous regions of MAGP-36 (bovine) and fibrinogen alpha
(residues 120-132) and beta residues 338-353 (human) are also
shown.  "( )" is used to designate an ambiguous residue; "."
denotes a non-conserved residue; and "!" denotes a tryptic 
cleavage site.

             $ (Y)F P(F)V  D L M V M  A N Q P M     AAAP-40  
122  T L K Q K  Y E L R V  D L E D F  E N N T A     MFAP-4 
     T L   L K  Y E L R V  D L E D F  E X N T A     MAGP-36
                    L R V  E L E D .  A . N . A     Hum Fib-a 

     G E ! Y Y  D F F Q Y  T X G M A  K E Y D G F Q AAAP-40
142  Y A K Y A  D F S I S  P N A V S  A E E D G     MFAP-4
     F A K Y A  D F S I S  P N A V S  A E E D G     MAGP-36

     Y T X G M  A K(I Y)A  G N A L M  D G A S G L M AAAP-40     
162  Y T L F V  A G F E D  G G A G D  S L S Y H     MFAP-4 
     Y T L Y V  S G F E D  G G A G D  S L T Y H     MAGP-36
     Y . I . V  . K   T A  G N A L                  Hum Fib-b 


                            TABLE II

     Alignment of experimentally determined sequences of AAAP-40 on
sequences from human vitronectin (VN, residues 230-240) and MFAP-4
(beginning at residue #34).  Alignments with other calcium-binding
proteins are shown, with the most highly conserved residues
highlighted in bold type: Calcium-binding myeloid-related protein
= CBP-M; Aggrecan (bovine); human fibrinogen beta (residues 144-
157 from calcium-binding domain) = Hum Fib-b.  

Q E L E K                        $ F E D G V L D P D Y P  AAAP-40
                                 R F E D G V L D P D Y P  VN
F C L Q Q P L D C D D I Y  A Q G Y Q S D G VYL I Y P S    MFAP-4
S E L Q L P L D E D D I Y  A Q G Y Q A D G VYL I   P S    MAGP-36
T E L . . . L . E . D V Y  . . . Y . . D                  CaBP-M  
        P . D E . D V Y                                   Aggrecan
S E L E K H Q L . . D . T                                 Hum Fib-b

                          REFERENCES
--------------------------------------------------------------------
A genetic basis for autoimmune manifestations 
in the abdominal aortic aneurysm (AAA) resides 
          in the MHC Class II Locus DR-beta-1

M. David Tilson, MD
Kathleen J. Ozsvath, MD
Hitoshi Hirose, MD
Shichao Xia, MD

From: Columbia University and the Department of
Surgery, St. Luke's/Roosevelt Hospital Center, New York

Corresponding author:

     Dr. Tilson
     Dept Surgery
     SLR Hospital Center
     1000 Tenth Avenue
     New York, NY  10019

This work was supported by the Special Fund for
Resident Research, the Department of Surgery, St.
Luke's/Roosevelt Hospital Center

Introduction

     Aortic aneurysms are a common disease of maturity,
among the ten leading causes of death in men and women
of ages 55-74.1  Despite all of the mortality and
suffering caused by this problem, research into the
causes of the abdominal aortic aneurysm (AAA) has been
neglected.  According to a recent search of the
computer resources at the National Institutes of Health
(NIH) on the subject of funded grants, the NIH is
spending something like 1.7 billion dollars a year on
AIDS, compared to a few hundred thousand dollars on
aneurysms; even though almost as many people die each
year from aortic aneurysms as from AIDS (about 25,000
versus 34,000).

     Autoimmunity has recently been implicated in the
pathogenesis of the AAA.  Immunoglobulin-G (IgG)
purified from the wall of AAA specimens is
immunoreactive with proteins of ~80 kDa and ~40 kDa
extractable from the aortic matrix and with matrix
elements that resemble elastin-associated microfibrils
in immunohistochemical preparations.   We have
purified and partially sequenced this putative
autoantigen.

     The notion that aneurysmal disease has features in
common with autoimmune diseases like rheumatoid
arthritis opens the way for many new research
approaches to the issues of treatment and prevention. 
If specific antibodies are detectable in serum, as
preliminary results suggest, it may be possible to
detect susceptibility to the disease before significant
aortic degeneration has occurred.  If tolerance for the
aortic autoantigen can be induced, it may be possible
to down-modulate progression of aortic degeneration (as
has been accomplished in rheumatoid arthritis).

Background

     It has been 18 years since Clifton reported an
instance of familial clustering of AAA and 15 years
since the first author  postulated a general genetic
basis for susceptibility.  Numerous subsequent studies
have added support to the genetic hypothesis.  
          The most recent
statistical model for the mode of inheritance predicts
a dominant gene.  

     Three candidate genes have been proposed.  1) A
mutated gene for the alpha chain of Type III Collagen
(COL3A1) was found to cosegregate with aneurysmal
disease in one family, but further studies revealed
only two additional positive families in another 50
evaluated.   2) A deficiency allele for alpha-1
antitrypsin (MZ) was found in five (11%) of 47
patients, so if this gene has etiological
significance in some individuals, it accounts for a
relatively small subset.  3) Our group found an
identical nucleotide substitution in the cDNA for
Tissue Inhibitor of Metalloproteinases (TIMP) in two of
six patients; but this finding turned out to be
trivial, since the polymorphism occured in the third
position of the codon.  Accordingly, the gene and/or
genes responsible for most cases of AAA remain to be
discovered.

Evidence for Autoimmunity in the AAA

     The walls of abdominal aortic aneurysms (AAA's)
contain abundant cells of chronic inflammation,
particularly in the 
adventitia.     We have also observed Russell
bodies in the AAA tissues, which are a feature of
autoimmune diseases like Hashimoto's thyroiditis.28 
Additional findings from our laboratory since that time
include 1) elevated levels of cytokines that activate
proteolysis in AAA tissue; 2) elevated content of
plasmin (the principal activator of the matrix
metallo-proteinases); 3) increased amounts of proteins
immunoreactive with antibodies against collagenase
(MMP-1), stromelysin (MMP-3), and gelatinase-B (MMP-
9).  In our original study of 1991 we noted
substantial increases in the IgG present in the walls
of aortic aneurysmal tissue by comparison to
atherosclerotic occlusive disease and control aortas28. 
This observation led to further work to identify the
putative autoantigen.

Purification of an autoantigen

     We purified IgG from aneurysm wall by affinity to
Protein-A and used it as a probe in immunohistochemical
preparations of normal aorta to determine whether the
autoantigen is an aortic matrix protein.2  Taking a
similar approach to the assignment of a molecular
weight to the putative autoantigen, we have used the
AAA IgG as a probe on Western immunoblots of soluble
proteins extracted from the normal aorta.2

     Amino acid sequencing of peptides from the
putative 80 kDa autoantigen3 led us to postulate that
our protein was the human homolog of a bovine aortic
protein discovered by Kobayashi, named microfibril-
associated glycoprotein-36 (MAGP-36).  MAGP-36
occurs in nature as a dimer, so we hypothesized that
our 80 kDa protein was also a dimer.  Further
extractions under chaotropic reducing conditions were
used to purify sufficient human protein with Mr ~40 kDa
to 1) Confirm that the 40 kDa species is also
immunoreactive with AAA IgG; and 2) Confirm that it is
closely related to the MAGP family by direct amino acid
sequencing.4 

     At first we called this 40 kDa protein "Aortic
Aneurysm Antigenic Protein-40" (AAAP-40).4  Since it
appears to be the third member of the MAGP family to be
discovered in man, the abbreviation MAGP-3 may be
preferred in the future.  AAAP-40 is homologous to two
other well-characterized microfibril-associated
glycoproteins: MAGP-3638 and MFAP-4.

Parallels with rheumatoid arthritis

     The association of rheumatoid arthritis (RA) and
Class II Major Histocompatibility Locus (MHC) DR4 was
first reported by Stastny in 1978.  However, it
became evident that DR4 was not always associated with
RA in different racial/ethnic populations, and the
reader is referred to a review article by Ollier and
Thomson which describes the state of the field
leading to the formulation of the hypothesis of the
shared epitope by Gregersen, Silver, and Winchester in
1987.  The hypothesis is that arthritogenic DR
molecules share a highly conserved sequence of amino
acids in their third hypervariable region (amino acid
positions 70-74).  

     In the context of a rather homogeneous population
seen by the arthritis group at the Mayo clinic,
evidence for the shared epitope seems compelling. 
Weynand and colleagues have reported that 98 of 102
(96%) patients express one of the major North American
disease-linked polymorphisms (*04, *0101, or *1402). 
Forty-seven patients carried a double dose of the
relevant sequence stretch.  Nodular disease was
expressed in 100% of the patients typed as HLA-
DRB1*04/04.  Patients with a double dose of the shared
sequence tend to have more severe manifestations of
rheumatoid disease.

A pilot study to test the notion of an association with
MHC Class II DR.

     As a pilot study we tissue typed blood specimens
from 25 patients with AAA's (some fresh and some
frozen).  Because of the demographics of the patient
population of our hospital system, we were fortunate to
have four Americans of color in the sample.  Since
AAA's are less common in African Americans than North
American Caucasians, these eight haplotypes were
particularly interesting.  It has been a working
hypothesis of the first author for some years that
Americans of color might have a double dose of the
susceptibility allele.  The results of allele frequency
analysis in these four patients compared with the
expected frequencies are presented in Table III.  The
expected frequencies were derived from the tables
published by the Eleventh International
Histocompatibility Workshop held in Yokohama, Japan in
November 1991.  The data in the Table  suggest that
patients with alleles for DR 2, 12 and 13 were detected
more often than expected by chance alone (p=.0003).

     Inspection of the amino acid sequences of the most
common alleles for these three DRB1 types revealed that
the residues that they have in common are at positions
31 and 47 of the second hypervariable region;
specifically, both residues are phenylalanines.  The
other DRB1 alleles that have phenylalanine at positions
31 and 47 are DR3 and DR 11.  Revisiting the data on
all 25 patients in the original series, 20 of 25 (80%)
have an allele with phenylalanine at positions 31 and
47; and 8 of 25 have a double dose of one of the
putative "aneurysmogenic" alleles.  Thus, there is now
a testable hypothesis to evaluate as we move forward to
look at additional patients prospectively.

Looking toward the future.

     The notion that AAA is an autoimmune disease of
maturity like rheumatoid arthritis offers many
opportunities for new knowledge, including 1) a better
understanding of etiological influences; 2) novel
diagnostic approaches to the detection of
susceptibility (DR) and disease activity (serum IgG
level) before the aorta has commenced to dilate; and,
3) therapeutic interventions (induction of tolerance)
that might change the natural history of the disease. 
In terms of etiology, we have found several instances
of molecular mimicry by common (and uncommon)
pathogens,  wherein the shared epitopes may provide
informative explanations not only for features of AAA
disease, but also for previously unexplained phenomena
related to the clinical manifestations of certain
infections.

TABLE

     Allele frequencies in four Americans of Color with
AAA (AOC-AAA, eight haplotypes) with the expected
frequencies for these alleles in North American Blacks
(NAB) as reported by the International Workshop in 1991
(Table 12, 132 haplotypes).43  The cumulative
probability of this result is p=.0003.  Probabilities
were calculated by Fisher's Exact Test for 2x2
Contingeny Tables.


Allele   Frequency in AOC-AAA  Frequency in NAB     p

DRB1*02      40%                      12%         .037
DRB1*12      30%                       5%         .023
DRB1*13      30%                      15%         .370